Analysis of product usage panel data

The aim of this thesis is to investigate approaches for analysing consumers' use of fast moving consumer goods over short periods of time. To date much research has been carried out to establish product purchase behaviour, but little is known about how products are used and by whom once they are brought. One wave of the European Personal Care Panel is used to investigate analysis approaches. Each wave of data is collected over a six month period and consists of around 2000 people from five European countries: France, Germany, Italy, Spain and the UK. Each panellist is asked to complete a diary and a background questionnaire.The diary data is assumed to have a hierarchical structure. Discrete time multilevel event history models have been used to analyse individuals' use of products over time. The probability of product use in a given hour conditional on a set of covariates is modelled allowing for random variation between people and days. The length of time since the product was last used and time of day have been incorporated in the models developed, along with person level covariates. Binary and multinormal models have been used to investigate single and joint product usage. These models have been estimated using the software package Mln (Rasbash et al, 1995). The incorporation of observations with left censoring has also been considered.